The Internet Just Became Majority AI-Generated: Welcome to the Slop Epoch

AI is taking over! Make a statement like THIS to let people know what side you are on

Synopsis: A new analysis of Common Crawl data suggests that as of May 2025, AI systems are now writing more web articles than humans. The internet has entered a “slop epoch” — a flood of generative-AI content that dilutes signal, amplifies noise, and forces every reader to play “spot the human.” Here’s how we know, why it matters, and what you can do to navigate this new age of digital dodgeball.

How We Know: Common Crawl, Graphite, and the AI Tipping Point

• The SEO firm Graphite analyzed a random sample of 65,000 English-language URLs drawn from the Common Crawl dataset. They focused on articles published between January 2020 and May 2025 and applied Surfer’s AI detection tool to each chunk of text. Graphite+1

• Their threshold: if 50% or more of an article is flagged as AI-written, label the article “AI-generated.” According to their findings, 52% of new articles as of May 2025 are AI-generated — over half of the web. Futurism+2Graphite+2

• That marks a tipping point: for the first time, AI seems to outproduce human authors across the open web, at least in the Graphite/Common Crawl sample. eWeek+3Futurism+3Graphite+3

• Caveats matter: detecting AI output is inherently noisy. Graphite reports a ~4.2% false positive rate (human content misclassified as AI) using Surfer’s detector. Graphite

• Additional research adds nuance: an academic analysis of linguistic signals (e.g. token patterns or “AI-style” phrase usage) estimates that at least ~30% of active web text is AI-derived, possibly trending toward 40%. arXiv

• Meanwhile, the opt-out movement is changing the playing field: many high-profile publishers are blocking AI crawlers or excluding their content from Common Crawl, which biases the measurements by excluding human-rich sites. Common Crawl+1

In short: while the statement “the internet just became majority AI-written” is bold, it's grounded in credible signals — albeit with serious measurement challenges.

Why It Matters: The Slop Epoch and the Crisis of Trust

If over half of what you read is generated or heavily aided by AI, that changes the rules of the game. Here are some of the biggest impacts:

1. Dilution of Signal & Rising Noise

• When volume is no longer scarce, quality becomes the differentiator. But “AI slop” is already a term for low-effort, filler content created for scale rather than meaning. Wikipedia

• Many AI-generated pieces lack original insight, deep reporting, or human judgment. They may repeat shallow patterns or hallucinate details.

• For content consumers, distinguishing signal from background noise becomes harder — every assertion now requires extra skepticism.

2. Search & Ranking Wars: Humans Versus Machines

• Search engines (especially Google) and AI assistants now face the challenge of curating content when much of the indexed pool is AI-generated.

• Graphite’s companion reports claim that 86% of top-ranked pages in Google Search remain human-written, compared to 14% AI — suggesting search may devalue pure AI slop. eWeek

• AI assistants (ChatGPT, Bard, Perplexity, etc.) also tend to prefer higher-trust sources, which may protect human authors to some degree — but the battlefield is shifting.

3. The Autophagous AI Loop

• Once AI models train on content that was itself generated by AI, they can begin feeding their own outputs back into the training loop.

• This raises risks: style collapse, echo-chamber reinforcement of errors, decline in diversity, and cascading hallucinations.

• Some scholars have warned of autophagous collapse, where models eventually “learn themselves into degenerate styles” if human signal is crowded out. arXiv+1

4. Editorial Control & the Power Shift

• If platforms or AI providers dominate distribution, independent creators and niche voices lose leverage.

• The “crawl layer” (whether your site gets indexed by crawling systems) becomes as critical as content SEO. Common Crawl+1

• Many publishers are pushing back — blocking AI crawlers, instituting paywalls, or demanding licensing from AI systems. Common Crawl+2arXiv+2

5. Epistemic Uncertainty & the Trust Crisis

• If an AI is writing vast swaths of what you read (including “news,” advice, opinion), how do you trust anything?

• Every reader becomes a detective, checking sources, dates, author credibility, and cross-referencing claims.

• In an age of misinformation, AI­-driven “pseudo-articles” or mistaken generative content could mislead at scale.

Can You Trust Anything Online Now? (Including This)

The short answer: trust cautiously.

• Always check source credibility: publisher reputation, author credentials, external references, and domain history.

• Be skeptical of blanket claims, statistics, or sweeping predictions.

• Prefer content that cites verifiable data (primary sources, studies, expert interviews).

• When possible, cross-check multiple trusted outlets.

• For critical information (health, finance, safety), rely on recognized authorities and peer-reviewed sources.

Even this article could itself be influenced (or challenged) by AI detection tools — transparency and accountability are now part of good writing.

AI is taking over! Make a statement like THIS to let people know what side you are on

What Readers & Creators Should Do Now

As Readers / Consumers

• Develop media literacy skills: learn to spot AI telltales (e.g. repetitive phrasing, weak reasoning, absence of original voice).

• Use detection tools (with caution) to flag suspicious content, but don’t treat them as perfect.

• Give feedback: report errors, flag misleading content, support writing that cites sources and shows human effort.

As Creators / Publishers

• Embrace hybrid authorship: use AI as an assistant or draft tool, but insist on human editing, unique insight, fact-checking, and narrative voice.

• Clearly label AI assistance (where used). Transparency is credibility currency.

• Build walls against AI misuse: deter content scrapers, use bot management, and enforce copyright / licensing.

• Focus on depth, originality, and authority — those will become sharper differentiators.

• Prioritize being part of the “crawlable web” (don’t block indexing completely) unless you negotiate terms with AI platforms.

• Experiment with watermarking, content authenticity frameworks, and new ownership models (e.g. Web3, federated publishing).

A New Epoch: AI Majority ≠ Human Irrelevance

We’ve entered a new era: the Slop Epoch, the point where generative AI content is no longer fringe but central. The internet isn’t dead yet — but its ecosystem is changing rapidly. The burden of trust now shifts more heavily to readers and creators.

Yes — that includes this article. It's your decision whether to believe me. But hopefully it gives you tools and perspective for navigating an internet where “who wrote that” has become one of the most urgent questions.